INTRODUCTION

BOLD signals reflect hemodynamic changes, and so are considered indirect indicators of neuronal activity. The measured extent of BOLD activations depends on the integrated contributions from the intrinsic spatial distribution of the neural activity involved, the effects of converting neural electrical activity to a spatial distribution of metabolic and hemodynamic changes that then affect MRI signals, and the effects of image acquisitions and reconstruction with limited resolution, and previous work has not adequately demonstrated what the relative contributions of these effects are in practice. Proper interpretations of fMRI studies rely on understanding the precise relationships between BOLD signal changes and their corresponding electrophysiological signatures. It is unknown to what extent BOLD signals correlate spatially with underlying neuronal activity, and the spatial relationships between stimulus-evoked BOLD signal changes and intrinsic resting-state functional connectivity of BOLD. To address these questions, we directly compared the spatial extents of stimulated activations and the profiles of inter-voxel resting-state correlations, between high resolution BOLD at 9.4T and local field potentials (LFPs) using two 49-channel microelectrode arrays, in functionally distinct somatosensory areas 3b and 1 in monkeys. We found the spatial profiles of BOLD and LFP were comparable, and the resting-state functional connectivity of BOLD was in agreement with that of LFP. Moreover, the BOLD responses at 9.4T were as focal as the measured underlying LFPs.

METHODS

FMRI: Stimulus-evoked (8 Hz vibrotactile stimulation of single digits) and resting state BOLD fMRI data (with 0.274 x 0.274 x 2mm³ resolution) were obtained from primary somatosensory areas 3b and 1 using a 9.4T MRI system and multi-shot echo-planar acquisitions. We estimated the spatial profiles of single digit BOLD response fields by computing the full width at half maximum (FWHM) of the fitted Gaussian point spread function (PSF). We also investigated the spatial extents of inter-areal (between areas 3b and 1) functional connectivity of resting-state fMRI signals. Electrophysiology: Two 7x7 multi-channel microelectrode arrays (98 channels in total, 400μm electrode spacing), were used to map spontaneous signal fluctuations in LFPs at rest and in response to 8 Hz vibrotactile stimulation of single digits in somatosensory areas 3b and 1 in anesthetized squirrel monkeys. We examined the spatial profiles of single digit LFP response fields by quantifying the signal power change and computing FWHMs. We estimated the spatial extents of inter-areal (between area 3b and area 1) functional connectivity by analyzing coherence characteristics of the spontaneous LFP signals at rest.

RESULTS

The variation in activation centers for repeated measurements by BOLD was less than 0.25 mm and was smaller for LFPs (~0.10mm), which were close to the BOLD in-plane resolution and smaller than the electrodes spacing, respectively. Activation maps of both modalities were co-registered within 0.10 mm accuracy. Figure 1 showed that in both modalities, the PSF of area 3b was more restricted along the medial-lateral direction, and stimulus responses in area 1 were less focal than those in area 3b. Figure 2 illustrated that in both modalities, the spatial extents of inter-voxel resting-state functional connectivity profiles were slightly wider than those of stimulated responses. There were no significant differences in PSF between LFP and very high resolution BOLD fMRI at 9.4T in both stimulus-evoked and resting-state conditions, and the resting-state functional connectivity of BOLD was in agreement with that of LFP.

DISCUSSION

Our results showed that the variation in activation centers for repeated measurements by BOLD and LFP was smaller than one voxel within each modality, and the separation of activation centers between modalities was very similar, indicating the stability and reproducibility of stimulus activation locations within and across two modalities. The PSFs of BOLD and LFP responses under stimulus conditions were around 1 mm, in which stimulus responses in area 3b were more focal in the medial-lateral direction than those in area 1. The PSFs of both BOLD and LFP correlation profiles in a resting-state condition were slightly wider than those of stimulus activations. Finally, we found the spatial agreement of resting-state functional connectivity between BOLD and LFP, and there were no significant differences in the dimensions of PSFs between two modalities. The strong similarities of spatial extents between BOLD and LFP are consistent with findings regarding the contribution of LFP activity to the fMRI signal¹, suggesting that BOLD fMRI at high resolution and high field faithfully reflects the underlying neuronal activity.

CONCLUSION

BOLD fMRI at high resolution and high field spatially faithfully reflects the underlying neuronal activity, which provides a viable tool for the investigation of cortical micro-organization at the meso-scopic modular level.

Acknowledgements

This work was supported by NIH grants NS078680 (JCG) and NS069909 (LMC). The authors gratefully acknowledge Fuxue Xin and George Wilson 3rd for the assistance with data collection and Chaohui Tang for the animal preparation.

References

1. Logothetis N.K., Pauls J., Augath M., Trinath T., Oeltermann A. Neurophysiological investigation of the basis of the fMRI signal. Nature 2001; 412: 150–157.